<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>The capsule network has good spatial recognition and has good accuracy in classification and recognition tasks. However, because of the dynamic routing algorithm in the capsule network, the training speed of the capsule network is slow. In order to make better use of the capsule network, reduce For its training cost, this paper proposes a capsule network based on the attention mechanism, and adds the CBAM attention module to the original capsule network to improve the network’s ability to extract information in the feature map channel and information in the feature map space, and improve the network’s learning ability, To reduce the number of network training, thereby reducing the cost of training. This paper conducts experiments based on the original neural network to verify the effectiveness and feasibility of adding the CBAM module to the capsule network. The final result is that the CBAM module can speed up the convergence speed of the capsule network by 50%.</p>
</abstract>

The capsule network has good spatial recognition and has good accuracy in classification and recognition tasks. However, because of the dynamic routing algorithm in the capsule network, the training speed of the capsule network is slow. In order to make better use of the capsule network, reduce For its training cost, this paper proposes a capsule network based on the attention mechanism, and adds the CBAM attention module to the original capsule network to improve the network’s ability to extract information in the feature map channel and information in the feature map space, and improve the network’s learning ability, To reduce the number of network training, thereby reducing the cost of training. This paper conducts experiments based on the original neural network to verify the effectiveness and feasibility of adding the CBAM module to the capsule network. The final result is that the CBAM module can speed up the convergence speed of the capsule network by 50%.

Research on Capsule Network Based on Attention Mechanism

International Journal of Advanced Network, Monitoring and Controls

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

{"article-title":"Research on Capsule Network Based on Attention Mechanism"}

The capsule network has good spatial recognition and has good accuracy in classification and recognition tasks. However, because of the dynamic routing...

Procedure	Routing algorithm
1:	procedure ROUTING $({\hat{u}}_{i ∣ j}, i, l)$
2:	for all capsule i in layer l and capsule j in layer (l + 1):b_ij ←0
3:	for r iterations do
4:	for all capsule i in layer l : c_i ← softmax (b_i )
5:	for all capsule j in layer (l +1) : s_j ← $\sum_{i} C_{i j} {\hat{u}}_{i \| j}$
6:	for all capsule j in layer (l +1) : v_j ← squash (s_j )
7:	for all capsule i in layer l and capsule j in layer (l + l)
8:	return v_j

Layer	Parameters
Conv	inputChannel:1; outPutChannel:256 kernel size= 9; stride=1
CBAM	Channel:256
PrimaryCaps	InputChannel:256;OutputCaps:3266output_dim:8;kernel_size:9,stride:2
MgitCaps	Inputeaps:3266;out_put_caps:10

Net Work Name	Routing Number	CBAM Module	Max Accuracy	First time
CapsNetl	1	False	99.70999908%	133
CapsNetl_CBAM	1	True	99.66999817%	100
CapsNet2	2	False	99.65000153%	111
CapsNet2_CBAM	2	True	99.61000061%	42
CapsNet3	3	False	99.69000244%	131
CapsNet3_CBAM	3	True	99.62002324%	82
CapsNet4	4	False	99.59999847%	141
CapsNet4_CBAM	4	True	99.55999756%	86
CapsNet5	5	False	99.65000153%	125

Research on Capsule Network Based on Attention Mechanism

Online veröffentlicht: 21. Feb. 2022

Seitenbereich: 1 - 8

DOI: https://doi.org/10.21307/ijanmc-2021-011

Schlüsselwörter
Component, Capsule Net, CBAM, Attention

© 2021 Yan Jiao et al; published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

j.ijanmc-2021-011.utab.001

NETWORK MODEL AND PARAMETERS

NETWORK MODEL ACCURACY

Research on Capsule Network Based on Attention Mechanism

Online veröffentlicht: 21. Feb. 2022

Seitenbereich: 1 - 8

DOI: https://doi.org/10.21307/ijanmc-2021-011

SchlüsselwörterComponent, Capsule Net, CBAM, Attention

© 2021 Yan Jiao et al; published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

j.ijanmc-2021-011.utab.001

NETWORK MODEL AND PARAMETERS

NETWORK MODEL ACCURACY

Schlüsselwörter
Component, Capsule Net, CBAM, Attention